HNHacker News
TopNewBestAskShowJobs

perrygeo

3,194 karma · joined April 28, 2014

submissionscomments
perrygeo··on The Gleam Programming Language
In a language that is otherwise as simple as it could possibly get away with (no `if`!), `use <-` initially feels like magic and somewhat out of place.

But take look at nested callback code, the pyramid of doom, and you see why it's pragmatically necessary. It's a brilliant design that incorporates just enough metaprogramming magic to make it ergonomic. The LSP even lets you convert back and forth between nested callback style and `use`, so you can strip away the magic in one code action if you need to unravel it.

perrygeo··on Don't fall into the anti-AI hype
LLMs tend to rise to the level of the complexity of the codebase. They are probabilistic pattern matching machines, after all. It's rare to have a 15 year old repo without significant complexity; is it possible that the reason LLMs have trouble with complex codebases is that the codebases are complex?

IMO it has nothing to do with LLMs. They just mirror the patterns they see - don't get upset when you don't like your own reflection! Software complexity is still bad. LLMs just shove it back in our face.

Implications: AI is always going to feel more effective on brand new codebases without any legacy weight. And less effective on "real" apps where the details matter.

The bias is strongly evident - you rarely hear anyone talking about how they vibe coded a coherent changeset to an existing repo.

perrygeo··on All my new code will be closed-source from now on
I've never thought of open source as something you can make money on directly. It's hard to see how it benefits an IC economically, besides getting some recognition and a sense of pride.

Open source has always felt explicitly like a benefit for companies.

- They get free code, buy vs build is irrelevant when you can just pip install.

- Systems become largely homogenized, thus contributors are replaceable.

- They get an established pool of workers who know the technology already, no training required.

- They get free labor from contributors outside their organization maintaining their dependencies for them in perpetuity.

It's a great deal for employers! Especially if they forbid their employees from contributing back! If you work out the game theory, there's literally no reason for a company to do anything but sit back and siphon the benefits for themselves.

This doesn't really change with LLMs, it just makes the end game much more explicit. The goal was always to capture the intellectual output of open source contributors for private profit. Always. Now that it's actually happening, who's really shocked?

perrygeo··on Ushikuvirus: Newly discovered virus may offer clues to the origin of eukaryotes
This. Mammals are (generally) K-selected species, meaning they invest heavily in raising their young. In the absence of natural pressures, mammals reproduce like crazy until they bump up against the environment's carrying capacity. Humans are not unique at all in our tendency to expand! It's just that we have opposable thumbs and language and tools to help us boost the carrying capacity.
perrygeo··on Show HN: 48-digit prime numbers every git commit
Even with git-prime reducing the address space by a few orders of magnitude, there's still (effectively) zero chance for collision. The difference between 10^-29 and 10^-27 isn't that great in practice.
perrygeo··on Tell HN: I write and ship code ~20–50x faster than I did 5 years ago
50x more code? Absolutely plausible. 50x more ideas implemented, or 50x better ideas? Doubtful. Generating code doesn't generate value.
perrygeo··on Show HN: Terminal UI for AWS
> low fidelity version of a browser UI?

That's the point. For me, with very few exceptions, modern web UI is steaming pile of dogshit - no consideration for user's attention, speed, or usability. TUI are extremely low fidelity; there's nowhere to hide all that enshitified cruft! Stripping the functionality down to its bare essence vs navigating a bespoke web UI with the design aesthetic of clown vomit. I can tell you which one is more productive for me.

perrygeo··on Supplementation with vitamin D3 reduced telomere attrition
This is exactly what I mean. You cite a single 12-yr old study and extrapolate its conclusion to extremes.

No mention of the Swedish cohort study (Lindqvist 2016) showing sun-avoiders had 2x mortality risk over 20 years. No mention of the dozens of ecological studies showing inverse relationships between UV and many cancers.

I could go on all day. You can't just paste one link and call it settled science.

perrygeo··on Supplementation with vitamin D3 reduced telomere attrition
As with anything health/nutrition related, the debate gets so comically two-sided with reductionist arguments talking past each other.

Get some sunlight but don't get a sunburn. It's not rocket science.

perrygeo··on Going immutable on macOS, using Nix-Darwin
Don't think of nix-darwin as an alternative to homebrew. Por que no los dos? You can run homebrew (declaratively!) via nix-darwin, and I think they complement each other quite well.

I stick to nixpkgs for most of the foundational system tools, lsps, build tools - basically anything that works in the terminal and is fully open source. For everything else (zoom, slack, vscode, firefox, etc) I use homebrew packages. The combo gives you options and is a strict superset of what you can do with homebrew or nixpkgs alone. Best of both worlds really.

The article I used to help set up, in case this helps anyone: https://davi.sh/blog/2024/01/nix-darwin/

perrygeo··on Why do teams keep explaining the same delays every sprint?
This is evidently a controversial take: if your projects are always delayed, maybe systematically track the things that slow you down and fix them?

In practice, this means instead of working around tech debt, you pay it down prior to working on it. It's faster than slogging through the same problems over and over. The key is you need to measure and focus on the ACTUAL things that slow you down. Lose a day to flakey tests? Fix the damn tests.

Eliminating the slow things is THE way to go fast - it's so obvious but very few teams actually do this. Instead they complain about all the problems and tech debt then act surprised when, lo and behold, those obstacles once again get in the way of progress. It's embarrassing to see a backlog of "these things slow down development!" tickets get systematically ignored as an explanatory factor, in favor of folk theories and politics.

perrygeo··on Stranger Things 5 Finale Ending Explained
For a show that takes ample time to over-explain every detail of the plot through repetition and analogy, I can't see how anyone would be confused by it. It's all there in the script a dozen times over.

I thought is was a good ending to a good show. It wasn't great, but it didn't need to be. They stuck the landing.

perrygeo··on Python numbers every programmer should know
> small int (0-256) cached

It's -5 to 256, and these have very tricky behavior for programmers that confuse identity and equality.

  >>> a = -5
  >>> b = -5
  >>> a is b
  True
  >>> a = -6
  >>> b = -6
  >>> a is b
  False
perrygeo··on Breaking Changes in Python-3.14
This is far from transparent. It's a massive change in behavior. Fork may be very hard to get right, but every python developer using multiprocessing has already paid that cost - and expects it to keep working!

With fork, you could pass objects that couldn't be pickled (lambdas, local functions, file handles, database connections). With forkserver, everything must be pickleable. That alone breaks thousands of repos of code.

You can no longer work with global module-level objects, so it fundamentally changes how scoping rules work.

It launches a server with some extra machinery at runtime - startup cost and hidden complexity just snuck its your app without you knowing.

forkserver may be technically a better choice. But that's irrelevant. Changing the default breaks existing code.

perrygeo··on The Gorman Paradox: Where Are All the AI-Generated Apps?
The link in the last paragraph provides some data to back up the claim. https://mikelovesrobots.substack.com/p/wheres-the-shovelware... - If the goal is to increase the rate of software production, there isn't much evidence that AI has moved the needle.

Sure, code gen is faster now. And the industry might finally be waking up to the fact that writing code is a small part of producing software. Getting infinitely faster at one step doesn't speed up the overall process. In fact, there's good evidence it that rapid code gen actually slows down other steps in the process like code review and QA.

perrygeo··on I tried Gleam for Advent of Code
The Gleam language, yes all of it, fits in a context window (https://tour.gleam.run/everything/)

I have similar concerns to you - how well a language works with LLMs is indeed an issue we have to consider. But why do you assume that it's the volume of training data that drives this advantage? Another assumption, equally if not more valid IMO, is that languages which have fewer, well-defined, simpler constructs are easier for LLMs to generate.

Languages with sprawling complexity, where edge cases dominate dev time, all but require PBs of training data to be feasible.

Languages that are simple (objectively), with a solid unwavering mental model, can match LLMs strengths - and completely leap-frog the competition in accurate code gen.

perrygeo··on Show HN: I audited 500 K8s pods. Java wastes ~48% RAM, Go ~18%
I'm not sure I would frame unused memory as "waste". At least not necessarily. If the system is operating below max capacity, that memory isn't wasted - it's a reserve capacity to handle surges. The behavior of the system under load (when it matters) might very well depend on that "wasted" memory.

You want to compare the _maximum_ memory under load - the high-water mark. If you combine Wozz with a benchmark to drive heavy traffic, you could more confidently say any unused memory was truly wasted.

perrygeo··on Programmers and software developers lost the plot on naming their tools
The cognitive tax is a real problem. Using random cutesy names, there is no obvious behavior boundary. Does authentication go in the SnufBux module? Or the Farfrumstable service? Who knows? Without obvious language clues to handle a new concept, any new functionality will get strewn across these internal boundaries. And why not - they have zero semantic meaning anyway! Sloppy names encourage sloppy programming.
perrygeo··on A “frozen” dictionary for Python
You're absolutely right: an "indirection data structure" is necessary. Freezing the data is the least interesting part - it doesn't give you any of the benefits typically associated with immutable data structures in functional languages. That's my point - Python is shipping a half solution that's being mistaken for a proper one.

You think Python developers are going to roll their own HAMT on top of frozendicts? Or are they just gonna make copies? Personally, I'd just use pyrsistent which seems to get it right.

perrygeo··on A “frozen” dictionary for Python
Python discovers immutable data, and gets it wrong. Frozendict is a blunt instrument - instead of a mutable free-for-all, it just locks the data for the entire lifecycle. Brace for the wave of code littered with deep copies, proudly proclaiming how functional it all is.

If you want real immutable data structures, not a cheap imitation, check out pyrsistent.

perrygeo··on If you're going to vibe code, why not do it in C?
> abstraction isn’t only for humans; it’s also for LLMs.

Bingo. LLMs are language models, not models of software systems. Everything gets translated through natural language! So the quality of the abstraction still matters: code that can be described well in plain language wins.

perrygeo··on Bag of words, have mercy on us
Predicting the next token is not at all the same thing as predicting the next action in a causal chain of actions. Not even close. One is model of language tokens, the other is a model of the physical world. You can come up with all sorts of predictions that can't be expressed cleanly in natural language. And plenty of things that parse cleanly from a language perspective but are unhinged in their description of empirical reality.
perrygeo··on Study links soybean oil to obesity
> “Soybean oil isn’t inherently evil,” Deol said. “But the quantities in which we consume it is triggering pathways our bodies didn’t evolve to handle.”

A rational response. Humans have always eaten a fair bit of omega 6. Just nowhere anywhere near _this_ much. Consumption has gone up at least 5x since the 1900s, in large part due to soybean/veg oil being in virtually every industrial food ingredient list. It's not evil, it's just everywhere and hard to avoid.

perrygeo··on Stopping bad guys from using my open source project (feedback wanted)
> I know my goal: shift the default in open source from “it’s free for anyone to use” to “please don’t use this if you’re evil”. I don’t just want to do this for my little project; I want to slowly change the discourse.

Good luck. Defining evil objectively is, of course, a challenge. But even with an unambiguous definition in hand, enforcing or detecting it is nigh impossible. Especially since the truly evil will simply lie, ignore the terms of your license, and use it anyway.

perrygeo··on Git 3.0 will use main as the default branch
Good. "master" implies a master/slave architecture which is straight-up technically wrong - and directly opposite git's true design! You couldn't pick a worse default. Even if it weren't tinged with racist colonial overtones, there is reason to change it for correctness alone.
perrygeo··on Homeschooling hits record numbers
Fully agree. The foundation of education is learning how the world actually is, not how we wish it would be.
perrygeo··on I made a downdetector for downdetector's downdetector's downdetector
I love the joke. The thing is, it's not so funny when companies actually try to do this on their own systems. I've never seen a 4x deep observability stack. But I have seen 3x, an entire project team dedicated to building a system that watches the watchers. It gets absurd.
perrygeo··on Why is Apache still popular even as Nginx has proven its mettle on performance?
A recent contract job of mine used Apache because they had written a critical component of their application as an Apache module. 10 years ago. There isn't much incentive to change it.

When people ask these questions, do they think people are actively choosing to use the old technology? That's a huge misunderstanding. It's not an apache vs nginx decision. It's do nothing vs spend precious time on a side quest to upgrade. Opportunity cost is your answer.

And re: performance, keep in mind that very few applications are limited by the speed of their HTTP server. You first look at your application servers, networks, disks, databases. If your app is truly HTTP-bound, well you're probably not still using Apache! IOW the people who NEED to upgrade from apache for performance reason already have. For the rest, there is no incentive.

perrygeo··on Immutable Software Deploys Using ZFS Jails on FreeBSD
Docker was the first viable containerization technology on Linux. Despite the 15 year late start vs FreeBSD Jails, it's certainly winning by the numbers.

But that has nothing to do with their respective UXs. It's a Linux vs FreeBSD signal.

perrygeo··on Exploring PostgreSQL 18's new UUIDv7 support
Is there an unavoidable tradeoff here? Keys that order nicely (auto-incrementing integers, UUIDv7) naturally leak information. Keys that are more secure (UUIDv4) can have performance problems because they have poor locality.

Or are there any random id generators that can compromise, remain sequential-ish without leaking exact timestamps and global ordering?

← PreviousPage 8 of 34Next →